Research On Energy-Saving Strategy Based On Coverage Method For Distributed Storage System

At present the popular Distributed Storage System(DSS)consists of inexpensive PC,which has advantages of high scalability,high availability,low cost,etc.However,plenty of idle nodes and massive amounts of data lead to a huge waste of energy.Therefore,it has enormous potential and great significance that conserving energy on this kind of DSS.As a general method for energy-saving,the energy-saving strategy based on coverage method is widely used by the DSS for energy-saving.Besides,most of energy-saving methods are based on Replicated Distributed Storage System(RDSS)rather than Erasure-Coded Distributed Storage System(ECDSS).However,with the advent of big data,many systems start to adopt erasure code for low storage cost and high fault-tolerant capability.Therefore,this paper first gives related theorem to prove feasibility of the coverage-based energy-saving method of ECDSS based on that of RDSS.Then based on the theorem,this paper gives several probability equations to analyze data availability after sleeping in different fault-tolerant mechanism that has the same redundancy and analyzes communication and storage overhead DSS that use different fault-tolerant mechanism through related equations.The results show that the DSS should take some measures to ensure availability with rising sleep rate.Besides,DSS will have lower communication and storage overhead when using erasure code of high rate and more fragments.Meanwhile,erasure code with low rate can obtain more saving compared to replication.Moreover,this paper point out the deficiency in computing overhead and proposes a coverage-based energy-saving method for RS code to improve computing overhead.It includes two algorithm called Redundancy First Sleep(RFS)and Data First Compensation(DFC)that correspond respectively with sleep node marking stage and data availability compensation stage.Then this paper analyzes DSS's computing overhead when adopting different erasure codes with different parameters.The results show the improved algorithm is feasible.Finally,this paper implements a DSS simulator based on discrete event to verify above theoretical analyses and evaluate the energy-saving method.The results show that using the energy-saving method in ECDSS can effectively reduce the communication and storage overhead.And the improved algorithms can effectively reduce the computing overhead of ECDSS to further improved effect of energy-saving and QoS.Meanwhile,it will not affect the advantages in communication and storage overhead.Moreover,the improvements will be better as the sleeping nodes grow up and the code rate reduces.